Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tell HN: ChatGPT is fantastic for finding and solving issues in logs

news.ycombinator.com

Post reply on HN
Print view View original

111–120 of 184 posts

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#111

Post by darkwater »

ohgodplsno wrote: post #98
Bootvis wrote: post #37

Earlier quoted context omitted.

In addition, for recipes, it’s just a better Google. If I do “Give me a concise recipe for X” it gives me one. No fluff, no ads. Just ingredients and steps. For example. I asked for pasta carbonara, concise and then even more concise. Final result: Quick Carbonara (4 servings) Ingredients: 12 oz pasta 4 eggs 1 cup cheese 8 oz bacon 4 garlic Salt, pepper Parsley (opt) Cook pasta, save water. Mix eggs, cheese. Fry baco…

Great if you want italians to get a stroke upon reading that you're putting bacon in carbonara, as well as your arteries dying at the thoughts of the sheer amount of pasta you've just made.

Ahahaha as an Italian that made me laugh! Anyway I wonder if it would better to ask ChatGPT to give the recipe in Italian (so it pulls memories from the Italian blogosphere corpus) and then translate it to English (or your language of choice)

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#113

Post by famouswaffles »

13years wrote: post #72
sharemywin wrote: post #58

Earlier quoted context omitted.

My understanding is the base model is pretty good about knowing whether it knows stuff or not. it's human feedback training that causes it to lose that signal.

Do you have any references? I know of the emergent deception problem that seems to be created through feedback. https://bounded-regret.ghost.io/emergent-deception-optimizat...

Base GPT-4 was highly calibrated. read open ai's technical paper.

also, this paper on gpt-4 performance of medical challenge problems confirmed the high calibration for medicine https://arxiv.org/abs/2303.13375

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#114

Post by mjrpes »

s3p wrote: post #103
akiselev wrote: post #75

Earlier quoted context omitted.

That log line ( with the four space at the front for HN formatting) is 40 tokens [1]. You can easily fit several hundred log lines with GPT4 8k context and with the incoming 32K context, you'll be able to fit close to a thousand log lines. That's a lot of context, especially if you can prefilter from relevant services, nodes, etc. or provide a multi-node trace [1] https://platform.openai.com/tokenizer

200 lines @ 40 tokens per line equates to 8,000 tokens. That costs $1.60. for one query.

For GPT-4 input cost is 3 cents per 1,000 tokens (8K context). So input cost for 8,000 tokens is 24 cents. Output is 6 cents per 1,000 tokens, but answer will hopefully be a lot shorter.

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#115

Post by d0mine »

q7xvh97o2pDhNrh wrote: post #62
ttul wrote: post #33

I used it to transform cryptic credit card statement items into company names, which then allowed me to query my Gmail archives for receipts and invoices from these vendors, automating a manual process of accounting backup discovery that is the bane of my very existence. I even got GPT-4 to assess whether an email likely relates to an invoice or payment so that I could limit the amount of noise extracted from my emai…

I keep seeing anecdotes like this, and I wonder: How do you feel about the privacy aspect of this? To do this, you had to feed your email into GPT-4, right?

Why do you think your email is private? Is your email provider more aligned with your interests or more secure than OpenAI? I doubt either Google or Microsoft care about your privacy (no difference).

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#116

Post by YeGoblynQueenne »

Bootvis wrote: post #37
darkwater wrote: post #30

Earlier quoted context omitted.

> Making meal plans: what is being done here? making a list of meals to eat each day of the week? isn't this just a question of thinking what one would like to eat? why is it easier to have the meals chosen by someone else? Do you have children? Meal planning can be a quite tedious and frustrating task if you want to cook at home, eat healthy, eat tasty, vary the dishes and have meals that kids will accept.

In addition, for recipes, it’s just a better Google. If I do “Give me a concise recipe for X” it gives me one. No fluff, no ads. Just ingredients and steps. For example. I asked for pasta carbonara, concise and then even more concise. Final result: Quick Carbonara (4 servings) Ingredients: 12 oz pasta 4 eggs 1 cup cheese 8 oz bacon 4 garlic Salt, pepper Parsley (opt) Cook pasta, save water. Mix eggs, cheese. Fry baco…

>> Combine, mix, season.

That "Combine" is doing all the work in there.

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#117

Post by simonw »

iso1631 wrote: post #83
simonw wrote: post #70

Earlier quoted context omitted.

I think there's a widely held misconception that anything you paste into GPT-4 will be used as raw training data by the model. Some people even seem to believe that it's learning continuously, so something you paste in could show up in an answer for another user a few minutes later. My mental model of how this works is somewhat different: - It takes months to train a model on raw data, and OpenAI train new ones (that…

> OpenAI DO NOT WANT your private data in their training data But they do want it. I can see many old chat logs. Data is a liability. Does "clear conversations" in chat.openai.com actually remove them? Or jst mark them as "deleted", but they remain in a database. I just did a data export, then a clear conversation, then another data export. The second export was empty, which seems suspiciously fast to me

I didn't say they weren't storing the data - obviously they have to store it to provide the (popular) chat log feature.

I said they didn't want it in the altar training data that they use for the pre-training phase of training future language models.

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#118

Post by 13years »

famouswaffles wrote: post #113
13years wrote: post #72

Earlier quoted context omitted.

Do you have any references? I know of the emergent deception problem that seems to be created through feedback. https://bounded-regret.ghost.io/emergent-deception-optimizat...

Base GPT-4 was highly calibrated. read open ai's technical paper. also, this paper on gpt-4 performance of medical challenge problems confirmed the high calibration for medicine https://arxiv.org/abs/2303.13375

Thanks, but I didn't find any details about performance of pre reinforcement training and after. Looking to understand more about the assertion that hallucinations are introduced by the reinforcement training.

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#119

Post by gumballindie »

ReaLNero wrote: post #66
dragonwriter wrote: post #32

Earlier quoted context omitted.

> Considering the fact that the RLHF for ChatGPT was done only in English but then worked just as well for every other language Does it work just as well for every other language, or does it work acceptably well for an important subset of other languages?

I speak a Slavic language spoken by ~2 million people, and I asked GPT-4 to tell me an old fable in my language. It did so fantastically, with no grammatical errors. I tried some more things -- it speaks it fluently, though admittedly not always idiomatically.

Meanwhile for actual uses cases it appears to be putting people at risk: https://restofworld.org/2023/ai-translation-errors-afghan-re...

Re: Tell HN: ChatGPT is fantastic for finding and solving issues in logs

#120

Post by justsomehnguy »

croes wrote: post #89
femto113 wrote: post #71

Earlier quoted context omitted.

You could try asking it to output the sed/awx/etc. commands needed to do the desired transformation reproducibly. If it's not yet good at that it will be soon.

The question is if you can understand the solution if it's complex. Just like the regex for valid email addresses, it may be correct but it's hard to understand it. I see it more as a tool for doing the tedious but simple work, if it's getting complex you get a hard time checking the correctness of the result.

Casual reminder what you shouldn't try to parse email addresses for a validity with regex, besides checking for '@' aaaaaaaand maybe for '@\w+?\.\w+?' if you are sure the code would only face the globally addressable addresses.
Post reply on HN